[Swift-devel] coaster status summary

[Ioan Raicu]

I looked around for some docs on the networking structure, but couldn't 
find anything.

There are several networks available on the BG/P: Torus, Tree, Barrier, 
RAS, 10Gig Ethernet.

Of all these, we are only using the Ethernet network, which allows us to 
communicate via TCP/IP (or potentially UDP/IP) between compute nodes and 
I/O nodes, or between compute nodes and login nodes.  For the rest of 
the discussion, we assume only Ethernet communication.  There is 1 I/O 
node per 64 compute nodes (what we call a P-SET), and the I/O node can 
only communicate with compute nodes that it manages within the same 
P-SET (the 64 nodes).  A compute node from one P-SET cannot directly 
communicate with another compute from a different P-SET.  This is 
primarily because compute nodes have private addresses (192.168.x.x), 
I/O nodes are the NAT between the public IP and the private IP, and the 
login nodes only have a public IP.  So, the compute nodes all have the 
same IP addresses, 192.168.x.x, and they repeat for every P-SET, and the 
I/O nodes handle their traffic in and out. 

Zhao, if you have any docs on the Ethernet network and the NAT that sits 
on the I/O node, can you please send it to the mailing list?

Ioan

Ben Clifford wrote:
>> it would matter on a machine such as
>> the BG/P, as there is a NAT inbetween the login nodes and the compute nodes.
>>     
>
> wierd. is there a description of that somewhere?
>
>   

-- 
===================================================
Ioan Raicu
Ph.D. Candidate
===================================================
Distributed Systems Laboratory
Computer Science Department
University of Chicago
1100 E. 58th Street, Ryerson Hall
Chicago, IL 60637
===================================================
Email: iraicu at cs.uchicago.edu
Web:   http://www.cs.uchicago.edu/~iraicu
http://dev.globus.org/wiki/Incubator/Falkon
http://dsl-wiki.cs.uchicago.edu/index.php/Main_Page
===================================================
===================================================